Southwest Research Institute (SwRI) develops drug discovery and chemical analysis software through multidisciplinary teams of computer scientists and chemists who support the pharmaceutical and biomedical sectors. We integrate artificial intelligence and machine learning into powerful tools that analyze chemical compounds for a variety of applications. Our machine learning in drug discovery solutions include:
- Structure-Based Drug Design – By incorporating machine learning in drug discovery, our Rhodum™ Molecular Docking Software rapidly interprets compounds and protein structures for Drug Discovery Research.
- Mass Spec Data Analysis – The R&D 100 Award-winning Highlight™ Non-Targeted Analysis System performs high-throughput mass spectrometry data analysis for Chemical Analysis Services and Pharmaceutical Development.
Machine Learning & Protein Docking
SwRI’s Rhodium virtual screening software integrates graphical processing and machine learning to scan hundreds of thousands of drug compounds per day. ML algorithms are constantly assessed in our environment and adapted into our virtual screening workflows. Ranging from artificial neural networks (ANNs), pairwise ranking (SVM) and clustering algorithms (kmeans), large-language models (LLMs), and convolutional neural networks (CNNs), graph neural networks (GNNs), we provide versatile and adaptable solutions for software-based drug discovery and platform development of emerging technologies.
The implementation of ML and AI in our Drug Discovery Research is not exhaustive:
| Architecture | Examples | Use Case |
| Artificial neural networks (ANNs) and convolutional neural networks (CNNs) | ReLu, MLP, arctan | Binary classification for discrete predictions of bioactivity |
| Pairwise ranking | Support vector machines (SVM), RankSVM and XGBoost | Rank ordering for predicting continuous distributions of bioactivity |
| Large language models (LLMs) | LLMs and natural language processing (NLPs) | Handling text-based chemical information such as SMILES, SMARTS, and SELFIES |
| Clustering algorithms | Kmeans and t-distributed stochastic neighbor embedding (t-SNE) | Data exploration of chemical space and atom type clusters of bioactivity datasets |
| Graph neural networks (GNNs) | Graph convolutional networks (GCN) and graph attention networks (GAN) | Pairwise chemical structure comparisons and consensus scoring metrics |
Learn more about our Computational Biomedicine or contact Jonathan Bohmann for more information.